BDD with Cucumber and Gherkin — genuine value or just an abstraction layer nobody asked for?

My new company uses Cucumber for all E2E tests. Feature files are written in Gherkin, step definitions in Java. After 2 months I'm struggling to see the value: - Product managers don't read or write the feature files (we write them for them) - The step definition layer adds indirection that makes debugging harder - Reusing steps sounds good but in practice steps are too context-specific to reuse Is this a failure of implementation (we're doing it wrong) or a fundamental problem with BDD for teams without non-technical stakeholders who actively engage with specs? I want to be fair to the approach before pushing for a change.